本文闡釋變偏EGARCH模型的經濟意義與性質,並嘗試在台灣股票市場上做進一步驗證,進行樣本內的模型參數估計與樣本外波動性預測能力評估。實證結果顯示:樣本內的估計結果支持對數化變幅(log range)是比對數化絕對報酬率(log absolute return)更有效率的波動性估計量。樣本外波動性預測能力的實證結果發現,變偏EGARCH族模型對於樣本外波動性的預測能力優於以報酬率為基礎的EGARCH模型,同時二因子波動模型的預測績效優於單因子模型。然而變幅EGARCH族模型對於樣本外波動性的預測能力僅達一個月,此結果異於與Brandt和Jones(2005)以S&P 500指數配適的結果,其原因可能是交易制度不同或市場結構差異所致。
This article describes the characteristics and implication of range-based EGARCH (REGARCH) model. The data of Taiwan Stock Index (TAIEX) are applied to examine the fitness of the REGARCH model and to test the out-of-sample volatility predictability of the REGARCH model. The empirical results show that log range is a more efficient estimator of volatility, and that the volatility predictability of REGARCH model is superior to that of the return-based EGARCH model. Besides, the predictability of the two-factor model is better than that of the single-factor one. However, we find that the volatility predictability is only for short horizons, which contradicts the conclusion of Brandt and Jones (2005) that predicting volatility with S&P 500 data can be as far as one year. Weimply that it is due to the difference of market systems.